-
AN ABSTRACT OF THE THESIS OF
Aveek Sarkar for the degree of Master of Science in Electrical
and Computer Engi-
neering presented on August 17, 1998.
Title: Radiation Effects in Compound Semiconductor
Heterostructure
Devices
Abstract approved.
S. Subramanian
Radiation studies were performed on compound semiconductor
heterostructure
devices. The objective was to understand the degradation
processes caused by the
exposure of these devices to radiation. Preliminary experiments
were focussed on
studying the degradation phenomenon in single heterojunctions
and single quantum
wells. It was found that ionizing radiation like gamma rays did
not have any sig-
nificant effect on the integrity of the III-V interface. Gamma
and neutron radiation
experiments were performed on heterostructure field effect
transistor structures to
study the degradation seen in the mobility and sheet carrier
concentration of the
carriers in the 2-DEG channel in these devices. In this
experiment it was found
that samples whose surface was protected with oxide passivation
suffered less dam-
age due to gamma radiation than the samples with unprotected
surfaces. However,
neutron radiation generated extensive damage in the 2-DEG
channel and this radi-
ation is believed to introduce charged defects into the spacer
and channel regions
in these devices. The fabrication process for heterostructure
bipolar transistors
(HBTs) was developed and high energy electron radiation studies
were performed
on GaAs/AlGaAs HBTs of different emitter areas and base widths.
The high and
low gain devices showed gain degradation of 40 and 14 %,
respectively, after an.elec-
tron fluence of about 9 x 1015 e/cm2. The factors that were
found to determine
the extent of degradation in the DC performance of the HBTs were
base width,
emitter contact area, surface condition, initial gain value and
the collector current
at which the gain was determined. Gamma radiation studies were
performed on
-
resonant tunnel diodes. It was found that the peak and valley
current levels were
almost unaffected even after a radiation dose of 40 MRad. The
voltage at which the
resonant peak occurs was found to increase and it is believed
that damage caused to
the contact regions is responsible for this shift.
-
@Copyright by Aveek Sarkar
August 17, 1998
All Rights Reserved
-
Radiation Effects in
Compound Semiconductor Heterostructure Devices
by
Aveek Sarkar
A THESIS
submitted to
Oregon State University
in partial fulfillment of the
requirements for the degree of
Master of Science
Completed August 17, 1998
Commencement June 1999
-
Master of Science thesis of Aveek Sarkar presented on August 17,
1998
APPROVED:
Major Professor, representing Electrical and Computer
Engineering
Head of Department o ectrical and Computer Engineering
Dean of Graduate S ool
I understand that my thesis will become part of the permanent
collection of Oregon
State University libraries. My signature below authorizes
release of my thesis to any
reader upon request.
Aveek Sarkar, Author
Redacted for privacy
Redacted for privacy
-
ACKNOWLEDGEMENTS
To Prof. Subramanian, whose hope of seeing this work culminate
into a doctoral
thesis I could not meet. I could not have wished for and got a
better teacher and a
finer gentleman to guide and assist me through a very tumultuous
period of my life.
For that and other things I cannot thank him enough.
To Profs. David J. Allstot, John F. Wager and Ben Lee for
advising and helping
me through my master's program.
To Profs. Tom Plant and Solomon Yim, for taking the time to be
on my defense
committee.
To Dr. Anirban Bandopadhyay, for sharing my faith in the HBT
fabrication
process and for having the patience to answer my questions while
I grappled with
the complex operational principles of these devices.
To Leon Ungier, who grew all the samples, kept the equipment
working and left
us to fend for ourselves. And to Marc Mims, who graciously put a
hiatus in his work
to get mine done. I wish him all success in his career and hope
that we both get the
opportunity to employ the knowledge we gained from working on
this project.
To the staff at the Radiation Center for carrying out all the
exposures.
Finally to Dr. Swati Sarkar, my mother and Souvik Sarkar, my
brother, for
letting me experience the joys and travails of graduate life
when they needed me by
their side.
This work is a part of the ongoing radiation study project under
the grant #
F49620-96-1-0173 from the Air Force Office of Scientific
Research.
-
TABLE OF CONTENTS
Page
1 Introduction 1
2 Physics of Heterostructure Devices 7
2.1 Physical properties of semiconductors 7
2.2 Heterojunctions 8
2.3 Carrier transport in heterostructure field effect
transistors 10
2.4 Heterostructure bipolar junction transistors 11
2.5 Single quantum wells and resonant tunnel diodes 14
3 Physics of Radiation Effects 17
3.1 Radioactivity and types of radiation 17
3.2 Radiation environments 18
3.3 Radiation damages in semiconductor devices 19
3.4 Measures of radiation effects 20
4 Radiation Study of III-V Interfaces 22
4.1 Experimental work 23
4.1.1 Sample preparation 23 4.1.2 Measurement set-up 24 4.1.3
Theoretical modelling 25
4.2 Results and discussion 26
4.2.1 Single heterojunction 26 4.2.2 Single quantum well 28
5 Radiation Study of Transport in HEMTs 31
5.1 Experimental work 32
5.1.1 Sample preparation 32
-
TABLE OF CONTENTS (CONTINUED)
Page
5.1.2 Measurement set-up 33 5.1.3 Measurements and data
collection 34 5.1.4 Theoretical modelling 35
5.2 Results and discussion 37
5.2.1 Gamma radiation 37 5.2.2 Neutron radiation 40 5.2.3
Summary 47
6 Radiation Study of Heterostructure Bipolar Transistors 48
6.1 Experimental Work 49
6.1.1 Individual processing steps 49 6.1.2 Mask set design 54
6.1.3 Fabrication sequence 59 6.1.4 Measurement configurations 69
6.1.5 Theoretical modelling 70
6.2 Electron irradiation experiment 71
6.2.1 Common emitter I-V characteristics 71 6.2.2 Gummel plots
78 6.2.3 Inverse Gummel plots 82 6.2.4 Current gain vs lc curves 82
6.2.5 Measurements on large area diodes 85
6.3 Discussion 87
7 Radiation Study of Resonant Tunnel Diodes 89
7.1 Description of devices 89
7.1.1 Structure 89 7.1.2 Layout of RTDs 90
7.2 Measurement procedure 92
7.3 Results 95
7.4 Discussion 99
-
TABLE OF CONTENTS (CONTINUED)
Page
8 Conclusions 102
8.1 Summary 102
8.2 Scope for future work 104
Bibliography 105
Appendix 109
-
LIST OF FIGURES
Figure Page
2.1. Energy band diagram in a heterojunction. The broken line in
the conduction band corresponds to a compositionally graded
junction to eliminate the conduction band discontinuity. . . 9
2.2. (a) A typical HFET structure, and (b) 2-DEG formation. .
11
2.3. Energy band diagram of a HBT under forward bias and the
associated current flow mechanisms. 12
2.4. A double barrier structure in a RTD: (a) at equilibrium and
(b) under biased conditions. 15
4.1. (a) Single heterojunction and (b) single quantum well
structures. 23
4.2. Measured carrier concentration profiles for the n-N
heterojunction. 27
4.3. Measured carrier concentration profile for the single
quantum well. 29
4.4. PL curves for the GaAs/InGaAs SQW after three doses of
radiation. Dose 0 = Pre-rad; Dose 1 = 1 MRad; Dose 2= 10 MRad; Dose
3 = 40 MRad. 30
5.1. Measurement set-up for Hall measurements. 33
5.2. Contact configuration in a typical Hall sample. 34
5.3. Measured (a) mobility, (b) carrier concentration values
obtained from an unpassivated modulation doped structure after four
doses of radiation 38
5.4. Measured (a) mobility, (b) carrier concentration values
obtained from a passivated modulation doped structure after four
doses of radiation. 39
5.5. (a) Delta-doped and (b) uniformly doped structures studied.
41
5.6. Normalized mobility and sheet carrier concentration values
for the uniformly doped sample (1-18-3-96) after neutron radiation.
41
5.7. Normalized mobility and sheet carrier concentration values
for the delta-doped sample (1-3-2-97) after neutron radiation. . .
42
-
LIST OF FIGURES (CONTINUED)
Figure Page
5.8. PL spectra for sample 1-29-10-93. Dose 1 = 4 x 1014 ii/cm2;
Dose 2 = 10 x 1014 lye 2;m Dose 3 = 36 x 1014 n/cm2. . 44
5.9. PL spectra for sample 1-21-10-96. Dose 1 = 4 x 1014 n/cm2;
Dose 2 = 10 x 1014 ilk 2;m Dose 3 = 36 x 1014 n/cm2. . 44
5.10. Defect introduction at 77 K versus neutron fluence. 45
6.1. Structures used for HBT fabrication. The Al mole fraction
in Al GaAs is 0 3 49
6.2. Etch rate calibration chart for GaAs /A1GaAs etch. 51
6.3. Layout of the masking levels on the plates. The numbers
denote the mask level numbers. The arrows indicate the orientation
of the particular level. 55
6.4. (a) The arrangement of different sized HBTs in a basic HBT
cell and (b) the arrangement of diodes and TLMs in a basic
Diode/TLM cell. 55
6.5. (a) Top view of a HBT structure detailing the mesas and the
contacts and (b) top view of a large area B-E diode structure. The
B-C diode had a similar shape with the collector contact
surrounding the smaller base contact. 56
6.6. Layout of the HBT and diode/TLM cells in a mask level.
57
6.7. Schematic cross section of the sample structure after
cleaving and lapping. 60
6.8. Schematic cross section of a HBT after the base mesa etch.
60
6.9. Schematic cross section of a HBT after the
emitter/collector contact deposition. 62
6.10. Schematic cross section of a HBT after the emitter mesa
etch and base contact deposition. 62
6.11. Schematic cross section of a HBT after passivation and
contact window opening. 64
-
LIST OF FIGURES (CONTINUED)
Figure Page
6.12. Schematic cross section of a HBT after the bonding pad
deposition. 64
6.13. Large area diodes and TLM patterns after contact
deposition. The bright metal regions are the base contacts
(Ti/Au-Zn/Au). 65
6.14. Photograph of the same sample after polyimide passivation.
. 65
6.15. A cell of HBTs after contact deposition. 66
6.16. Photograph of the same cell after polyimide passivation.
66
6.17. SEM picture of a one block of HBTs, diodes and TLMs.
67
6.18. SEM picture of a cell of HBTs with the bonding pads
(Ti/Au). 67
6.19. SEM picture of a single HBT. 68
6.20. SEM picture of a contact window with the metal finger
discernible. The emitter mesa appears in a darker hue compared to
the base layer. 68
6.21. Measurement configurations used. 69
6.22. lc vs VCE characteristics for D1H1 (passivated, 600 A
base) at IB---=10 ,u,A and 113=50 pA 73
6.23. lc vs VCE characteristics for D5H7 (passivated, 1000 A
base) at IB=10 ,uA and IB=50 pA 73
6.24. lc vs VCE characteristics for D2H7 (passivated, 1000 A
base) at IB=10 ,uA and IB=50 pA. 74
6.25. lc vs VCE characteristics for D2H5 (unpassivated, 1000 A
base) at IB=10 ,uA and IB=50 pA. 74
6.26. Base TLM measurement curves and the reverse saturation
current curves for the BE and BC diodes for the 1000 A base HBT
D2H5. 75
6.27. 1/i3 vs electron fluence for D1H1 (passivated, 600 A base)
at IB=10, 30 and 50 pA. 76
6.28. 1/0 vs electron fluence for D5H7 (passivated, 1000 A base)
at IB=10, 30 and 50 pA. 76
-
LIST OF FIGURES (CONTINUED)
Figure Page
6.29. 1/0 vs electron fluence for D2H7 (passivated, 1000 A base)
at IB=10, 30 and 50 ptA 77
6.30. 1/0 vs electron fluence for D2H5 (unpassivated, 1000 A
base) at IB=10, 30 and 50 itA 77
6.31. Gummel plots for D1H1 (passivated, 600 A base) before and
after radiation. 80
6.32. Gummel plots for D5H7 (passivated, 1000 A base) before and
after radiation. 80
6.33. Gummel plots for D2H7 (passivated, 1000 A base) before and
after radiation. 81
6.34. Gummel plots for D2H5 (unpassivated, 1000 A base) before
and after radiation. 81
6.35. HFE vs Ic curves for D1H1 (passivated, and after
radiation.
600 A base) before 83
6.36. HFE vs Ic curves for D5H7 (passivated,1000 A base) before
and after radiation. 83
6.37. HFE vs Ic curves for D2H7 (passivated, 1000A base) before
and after radiation. 84
6.38. HFE vs Ic curves for D2H5 (unpassivated, 1000 A base)
before and after radiation. 84
6.39. Carrier concentration vs depletion width for a large area
BE diode (D5BE3). 86
6.40. Reverse saturation current in a large area BE diode
(D5BE3). 86
7.1. Structure used to fabricate the RTDs. 90
7.2. Layout of RTDs on the die. 91
7.3. Arrangement of contacts in a RTD cell. 92
7.4. Probe placement on probe pads. 93
7.5. Characteristics for device B3A4R9. 94
-
LIST OF FIGURES (CONTINUED)
Figure Page
7.6. Change data for Vp in block 1. 97
7.7. Change data for Vp in block 3. 98
7.8. Change data for Vp in block 2. 99
7.9. Change data for Vp for all the 48 devices. ....... .
100
-
LIST OF TABLES
Table Page
4.1. Results of Kroemer's analysis after three doses of
radiation. 28
5.1. Typical pre-radiation values for the neutron samples.
40
5.2. Peak positions and FWHM values for samples 1-29-10-93 and
1- 21- 10 -96. 43
6.1. Feature (contact/mesa) sizes of the HBTs fabricated (sizes
in Ams). The HBT numbers are with reference to fig. 6.4. . . 58
6.2. Change in /3 after a dose of 9.2 x 1015 e /cm2. 72
6.3. Measured and predicted values of the HFE vs Ic power
coefficient at low values of IC. The radiated values are after a
dose of 9.2 x 1015 e /cm2. 85
7.1. Measured values of the four parameters for device B3A4R9
for different doses of radiation. 96
7.2. Relative change (w.r.t. un-radiated) in the measured
parameters for device B3A4R9 for different doses of radiation. . .
96
-
Radiation Effects in Compound Semiconductor Heterostructure
Devices
Chapter 1 Introduction
Fabrication of semiconductor devices and circuits requires
materials that have to
be prepared and engineered carefully. These devices and circuits
are susceptible
to failure from damage and defects that affect the purity and
crystalline order of
these materials. For semiconductor devices to operate reliably
and in a predictable
manner it is necessary to identify and control all known and
potential fault generating
mechanisms that arise during the fabrication and use of these
devices. Operation of
semiconductor devices under special conditions like deep space
and nuclear weapons
environments can cause various types of damage in the devices
depending upon the
type and energy of the nuclear radiation. The extent of these
different types of
damage has to be quantified so that the design and processing of
devices and circuits
can be formulated in a manner that can incorporate the
tolerances necessary to
offset such damage inducing mechanisms. The operation of space
crafts and satellites
depend upon the proper functioning of their on-board electronic
equipment. Thus,
the reliability issues arising from radiation exposure are very
important for circuits
used in such applications [1, 2, 3].
Energetic particles present in a radiation environment while
passing through a
material lose their energy by interacting with the atoms of the
material and through
different scattering mechanisms. These interactions result in
two types of damage in
the material: (1) atomic displacements (called displacement
damage) and (2) charge
build-up due to ionization (called ionization damage). The
extent of these damages
is a function of the material, the type of radiation and its
energy.
Radiation effects in silicon based devices like bipolar junction
transistors (BJTs)
and metal oxide semiconductor field effect transistors (MOSFETs)
have been exten
sively studied and modelled. In BJTs gain degradation and
increase in the leakage
-
2
currents are the primary effects of radiation. Displacement
damage increases the
number of recombination centers in the base region which
decreases the minority
carrier lifetime. Also damage to the surface and the passivation
layers increases
other base current components. The extent of these damages which
degrade the
gain of the transistor, depends upon the base doping profile,
base thickness and the
quality of the surface. Another effect that is seen is an
increase in the collector-
emitter saturation voltage (VcE,,,at) Low values of VcE,sat are
necessary for the
proper operation of switching power transistors and thus,
radiation induced increase
in VcE,sat can prove deleterious to their performance. Changes
in the minority car
rier lifetime, gain and the effective doping profile can be
related to the amount of
radiation through respective damage coefficients [1, 4, 5, 6].
These relations hold for
different types of radiation since the form of degradation is
similar. The factors that
determine the damage coefficients are the type of material,
doping level and profile
in the device layers, pre-irradiation gain value, type of
radiation and its energy. In
integrated circuits employing BJTs possible effects of nuclear
radiation are changes
in the biasing voltage and current levels. Exposure to ionizing
radiation can result
in parasitic photocurrent generation that can induce latchup
[6].
Radiation induced charge trapping in the oxide (or dielectric)
layer is the primary
cause of performance degradation in MOSFETs. The amount of
charge trapping
depends upon the electric field across the oxide during
radiation, the oxide thickness
and the material quality of the oxide which in turn depends upon
the processing
conditions during and after the oxide growth [9]. Atomic
displacements at the silicon-
silicon dioxide interface result in an increase in the density
of interface states. These
two effects: (1) generation of oxide trapped charge and (2)
increase in the density of
interface states affect the threshold voltage of the transistor.
Extensive modelling has
been done to correlate the shift in the threshold voltage with
radiation. Also charge
trapping in the field oxide in regions lying over the edge of
the source and drain
implants tends to cause inversion that shunts the main channel
region. Annealing,
both at elevated and at room temperatures, have been found to
ameliorate some of
the damages [7, 8, 9].
-
3
Devices fabricated using compound semiconductor materials (like
GaAs, Al-
GaAs, In GaAs) are finding increasing application as these
materials have several
attractive properties not available in silicon. Carrier mobility
in compound semicon
ductor devices are far superior to those achievable in elemental
semiconductors. Some
of the devices fabricated using III-V materials are metal
semiconductor field effect
transistors (MESFETs), high electron mobility transistors
(HEMTs), heterostruc
ture bipolar junction transistors (HBTs) and resonant tunnel
diodes (RTDs). These
devices are being used in high speed and microwave/millimeter
wave circuits. III-V
materials like GaAs have been found to be intrinsically more
radiation hard. So,
these devices are particularly attractive for use in areas where
stringent radiation
requirements have to be met. Radiation studies have been
performed on bulk GaAs
and on MESFETs but not much effort has been directed towards
studying radiation
effects in HEMTs, HBTs and RTDs.
The damage created by radiation in bulk GaAs manifests itself
primarily through
carrier removal which is a consequence of atomic displacements
in the material. Elec
tron irradiation studies have pointed toward the generation of
randomly distributed
vacancy-interstitial pairs while neutron irradiation results in
the formation of small
regions of intense damage. These defect clusters act as
scattering and trapping cen
ters for the free carriers whose mobility degrades in the
presence of these scattering
mechanisms. Transport studies have shown that the conductivity
type remains the
same indicating a deep level trap introduction process [10, 11,
12]. The carrier re
moval rate and mobility degradation have been related to the
particle fluence by,
N = No(1 fin On) (1.1)
1 1On) (1.2), (1 +
Po
Here, N(No) and p,(pto) are the doping levels and mobility
values after(before) radi
ation, respectively. The damage coefficients for the doping
level and mobility value
change are given by /37, and respectively. The particle flux of
the radiation is
given by On. The carrier removal rate has been found to be
dependent on the dop
ing profile of a material. Carrier removal is more severe in ion
implanted materials
-
4
with their non-uniform doping profiles as compared to materials
with more uniform
doping profiles obtained by diffusion [13].
In GaAs MESFETs radiation induced damage decreases the gate
pinch-off volt
age, the transconductance and the saturation current. The source
and drain resis
tances are found to increase after radiation exposure [14].
These effects have been
modelled through carrier removal and mobility degradation in the
channel region
[13]. In Schottky diodes the reverse bias leakage current
increases after radiation
[15]. Small changes in the surface state densities result in
slight increase in the for
ward bias current level. The barrier height which is primarily
determined by the
high density of surface states in GaAs has been found to remain
unaffected [16].
Radiation studies in HEMTs have encompassed the effects of gamma
, low energy
electron, neutron and helium ion radiation [17, 18, 19]. However
the focus of these
studies has been to explain the degradation in the device
characteristics like threshold
voltage and drain current. Not much effort has been directed
towards understanding
the physical mechanisms behind the degradation process since it
is difficult to isolate
the effects seen on intrinsic device parameters like mobility
from the effects seen on
extrinsic device parameters like drain and source series
resistances obtained from the
device characteristics. Some of the changes that have been found
to take place are
(1) shift in the threshold voltage due to electron trapping in
the buffer layer of the
HEMT, (2) change in the depletion layer width and (3) shift in
the Fermi level. The
last two changes affect the sheet carrier concentration in the
channel. The drain
current has been found to decrease by about 50 % at neutron
fluence levels of 3 x
1015 n/cm2 and it vanishes altogether after a fluence of 1016
n/cm2 [17]. Mobility and
sheet carrier concentration of the carriers in the 2-DEG have
been found to decrease
linearly with the particle fluence. The drain and source
resistances also increase with
radiation [18].
GaAs HBT technology with its superior performance levels in
terms of opera
tional frequency, power consumption, gain-bandwidth product, and
1/f noise provide
several advantages over advanced silicon bipolar junction and
GaAs MESFET tran
sistors. These features make GaAs HBTs likely components in
future high speed
-
5
circuits. However, radiation studies on HBTs have been very
limited and these too
have been restricted to neutron irradiation.
The nature and the extent of damages caused by radiation in a
HBT differs from
the damages suffered by a silicon BJT because of the presence of
a heterojunction in
the HBT. The degradation in the current gain (/3) in HBTs has
been found to be less
compared to Si-BJTs. Schrantz et al [20] reported a current gain
degradation of 60 %
and 96 % in HBTs and BJTs, respectively, after a neutron fluence
of 1015 n/cm2. In
HBTs the decrease in the values of the current gain has been
attributed to the degra
dation of the base-emitter (BE) heterojunction which increases
the recombination in
the BE space charge region. This increases the space charge
recombination current
component of the base current and raises the base current
ideality factor. Decrease
in the minority carrier lifetime has been found to have less
influence on 0 degradation
in GaAs /AIGaAs HBTs [20]. Gummel plots have indicated that for
high-0 devices
the base current rises both in the low and high VBE regions
while for low-0 devices
the base current remains unaffected in the high VBE region. Low
/3 devices have
been found to undergo less degradation compared to high 0
devices. This has been
explained by the fact that in low 0 devices surface
recombination currents play a
dominant role and these currents are supposed to be relatively
radiation insensitive.
Song et al [21] have reported a current gain degradation of 25 %
and 7 % in their
high and low 13 devices, respectively, after a neutron fluence
of 1.3 x 1014 n/cm2.
Gummel plots have shown the collector current to remain almost
unaffected in both
low and high /3 devices. Current gain degradation has been found
to be more severe
when (1) the initial value of /3 was high, (2) the collector
current level was low and
(3) the emitter area was large [20, 21].
The objective of this research is to develop a systematic
understanding of the
radiation effects in some III-V devices and quantum well
structures. Preliminary
experiments were focussed on studying radiation effects in
single heterojunction and
single quantum well structures since these form the building
blocks of the devices
subsequently studied. An understanding of the degradation
mechanisms in these
structures is necessary to explain the damage seen in the
devices that use them.
-
6
High carrier mobility values in the 2-DEG channel give HEMTs an
edge over other
devices that employ a channel for current conduction. Transport
studies were per
formed on HEMTs to quantify and explain the degradation in the
2-DEG channel.
This study differed from the reported studies in that it
employed direct measurement
of the intrinsic parameters like mobility and sheet carrier
concentration which deter
mine the performance of these transistors. Radiation effects on
HBTs had been the
primary thrust area of this work since it designed and developed
the HBT fabrication
process for the first time at Oregon State University. High
energy electron radiation
studies on HBTs showed that the degradation is similar in nature
to the neutron
induced degradation reported in the earlier studies. Gamma
radiation studies were
performed on RTDs which showed that these devices are
considerably immune to
gamma radiation damage.
The organization of this thesis is as follows. Chapter 2
explains briefly the
principle of operation of the devices that were studied in this
work. Chapter 3 covers
some aspects of radiation physics that is of relevance to this
research. Chapter 4
covers the gamma radiation studies performed on single
heterojunction and single
quantum well structures while chapter 5 details the gamma and
neutron radiation
studies performed in characterizing 2-DEG transport in HEMTs.
Chapters 6 and 7
cover the radiation studies on HBTs and RTDs, respectively. The
different radiation
sources that were used are described in the Appendix.
-
7
Chapter 2 Physics of Heterostructure Devices
The structure of a semiconductor device and its physics of
operation determines the
effects of radiation damage on it. An understanding of the
principle of operation
of a device is therefore necessary to gain physical insight into
the damages suffered
by it under radiation. Towards this aim, this chapter starts
with a brief discussion
about semiconductors in general and III-V materials in
particular. This knowledge is
required to appreciate the difference between elemental and
compound semiconduc
tor devices. The concept of a heterojunction and a single
quantum well is explained
and the devices that employ these structures in their operation
are subsequently
covered. High electron mobility transistors (HEMTs),
heterostructure bipolar tran
sistors (HBTs) and resonant tunnel diodes (RTDs) are the
heterostructure devices
that were studied in this research. These devices possess some
unique capabilities
that render them the potential of being the building blocks of
extremely fast circuits
in the near future. Since it is beyond the scope of this chapter
to develop the full
details of the physics of operation of these complex devices,
the explanations cover
only the parts that are germane to this work.
2.1 Physical properties of semiconductors
Elemental semiconductors like silicon are group IV elements in
which each atom
shares its four valence electrons with four other atoms to form
a tetrahedral struc
ture. When two silicon atoms are replaced by a gallium (group
III) and an arsenic
atom (group V), a variation of the diamond structure in which
each gallium atom
is surrounded by four arsenic atoms and vice-versa results.
Since a gallium and an
arsenic atom have in their valence shells three and five
electrons, respectively, chemi
-
8
cal bond formation between them through the sharing of their
valence shell electrons
results in an uneven distribution of charge. An unit cell of the
crystal then has eight
electrons with a polarization of charge between the two atoms.
Thus, compound
semiconductor devices have mixed covalent and ionic bonding.
The interactions between the outermost shell electrons of
neighboring atoms in
the crystal result in the formation of conduction and valence
bands separated by
a energy gap called the band gap (b.g.). When sufficient energy
(greater than the
b.g.) is imparted to the electrons in the valence band they can
cross over to the
conduction band and are then able to contribute to a current.
The excitation of an
electron from the valence band to the conduction band results in
the formation of
a positively charged hole in the valence band. This in turn
contributes to a current
flow due to the motion of the positive charges in the direction
opposite to that of
the electrons. Doping of semiconductors involves introducing
impurities that have
either more or less electrons than the atoms they replace in a
lattice site. Doping
results in the introduction of energy states within the b.g.
that are either close to
the conduction band or to the valence band and can contribute
electrons or holes
to the respective bands. The former results in a n-type material
and the latter a
p-type material. In an n-type material, higher concentration of
electrons make them
the majority carriers of current while holes form the minority
carriers. In a p-type
material the reverse is true. Commonly used n-type dopants for
GaAs are silicon,
tin, and sulphur while beryllium, zinc and carbon serve as
p-type dopants [22, 23].
2.2 Heterojunctions
A heterojunction results when two semiconductors with different
band gaps are
placed in contact. Close lattice match between the two materials
is required to
obtain a good interface free from defects. Compound alloy
semiconductors (GaAs,
AlsGa(i_x)As, InxGa(i_x)As, etc) have different band gaps
depending upon their
composition. Fig. 2.1 shows the energy band diagram of a
heterojunction that re
sults at equilibrium when a p-type GaAs layer (small b.g.) is
placed in contact with
-
9
p-GaAs N- A1GaAs E(x)
Evac
Ec
Ef Ev
Ev
0 xSCR
Figure 2.1. Energy band diagram in a heterojunction. The broken
line in the conduction band corresponds to a compositionally graded
junction to eliminate the conduction band discontinuity.
a N-type A1GaAs layer (wide b.g.). By convention, in a
heterojunction the conduc
tivity type of the smaller b.g. material is noted in lower case
letters while the same
for the larger b.g. material is noted in upper case letters. The
heterojunction can
be an isotype heterojunction in which both layers are of the
same conductivity type
(n-N, p-P) or an anisotype heterojunction (p-N, n-P, etc).
The difference in the band gaps of the two layers generates
discontinuities in
the conduction and valence bands at the interface. Using the
electron affinity rule
the conduction band discontinuity is given by, AE, = Xp-XN where
XP and XN are
the electron affinities of the p and N-type semiconductors,
respectively. The valence
band discontinuity is given by AK, = AEg-AE,. Diffusion of
excess electrons from
the N-side to the p-side and vice versa, results in the
formation of a space charge
region (SCR) at the interface. This charge polarization
generates an electric field
-
10
that cancels the diffusive flow of carriers so that at
equilibrium no net current flows
through the junction.
Application of a positive potential to the p-side results in a
forward bias (FB)
condition while application of negative potential produces a
reverse bias (RB) con
dition. When forward biased, the barrier to the injection of
carriers gets lowered
causing excess minority carriers to appear at the edges of the
SCR on both sides.
This generates a net diffusive current flow through the
junction. When reverse biased
the barrier to diffusive current flow increases and only a
junction leakage current flow
is possible due to the electric field across the junction. Also
generation and recom
bination processes within the SCR add to these currents during
reverse and forward
bias, respectively. The net current flow is represented as,
J = Ji(exp(qVImkT) 1) + J2(exp(qVInkT) 1) (2.1)
Here m and n are the ideality factors for the diffusion-drift
component and
generation-recombination components of the junction current.
2.3 Carrier transport in heterostructure field effect
transistors
Fig. 2.2 (a) depicts a typical heterostructure field effect
transistor (HFET) structure.
This device employs the principle of establishing a conducting
path (channel) between
two electrically isolated contacts (source and drain) based on
some conditions (bias
to the gate). An HFET differs from other devices using a similar
concept in that the
channel carriers in it are separated physically from their
donors. In the structure
shown electrons from the highly doped supply layer diffuse to
the undoped GaAs
layer and accumulate in the triangular potential well formed at
the interface due
to the band gap difference (fig. 2.2 (b)). This electron
concentration is called a
2-Dimensional Electron Gas (2-DEG) since the motion of the
electrons inside the
potential well is quantized in the direction perpendicular to
the interface. Self-
consistent solutions of Poisson's and Schrodinger's equations
give the sheet carrier
concentration in the 2-DEG. The concentration of carriers in the
2-DEG can be
-
11
Source Gate Drain
Al GaAs GaAs
2-DEG
(a) (b)
Figure 2.2. (a) A typical HFET structure, and (b) 2-DEG
formation.
varied by the application of bias to a Schottky barrier gate on
top of the supply
layer. Due to the separation of the carriers from their charged
donors, they suffer very
little Coulomb scattering. Also the high concentration of the
carriers in the channel
generates a screening effect that reduces the amount of
scattering they experience.
This results in very high mobility values of the carriers in the
channel. These devices
can therefore operate at very high speeds [24].
2.4 Heterostructure bipolar junction transistors
A heterojunction bipolar junction transistor (HBT) is a
variation of a bipolar junction
transistor that employs heterojunctions to achieve superior
performance levels. An
HBT consists of three regions: emitter (E), base (B) and
collector (C) and can be
visualized as a back-to-back combination of two diodes with the
base as the common
region between them. The basic principle of operation is that
the carriers injected
by one junction are collected by the other junction. The modes
in which an HBT
can operate are (1) forward active, (2) saturation, (3) cut-off,
and (4) reverse active.
These are determined by the bias conditions of the EB and BC
junctions.
-
12
EB SCR BC SCR B
Isr
Ec E Ef In
igr B
Iscr Ip
BC Ip
Emitter Base Collector
Figure 2.3. Energy band diagram of a HBT under forward bias and
the associated current flow mechanisms.
Fig. 2.3 shows the energy band diagrams of a N-p-n HBT in the
forward active
mode and the different currents flowing in the device in this
mode. In this device
the BE junction has been compositionally graded to eliminate the
spike and notch
discontinuity present in abrupt junctions. The emitter, base and
collector currents
in the forward active mode are given by,
IE Ip (2.2)
1- sBer ibBr IBCIB = IpB (2.3)
IC= InC ipBC (2.4)
Here, InE is the electron current from the emitter into the
base, Ip is the hole
injection current from the base into the emitter, LE, is the
space charge region re
combination current, IsE,. is the surface recombination current,
Ib is the quasi-neutral
region recombination current, IpBc is the generation current of
the BC junction and
Inc is the electron current reaching the BC junction.
-
13
The parameters characterizing the performance of an HBT are:
Emitter injection efficiency: IE
/7f /2;3 (2.5)
Common emitter current gain: /c Is (2.6)
It is desirable to have values of 7 very close to 1. For that
purpose the values of
IpB have to be much smaller than I. Since the BE junction is a
heterojunction
while the BC junction is a homojunction, the barrier seen by the
holes from the base
is more than that seen by the electrons from the emitter (fig.
2.3). In fact for a
graded junction, this barrier is equal to the band gap
difference between the emitter
and base materials. Since the excess carrier injection across
the junction decreases
exponentially with increasing barrier height, hole injection
from the base is very
small. To achieve a similar result in a bipolar junction
transistor, the emitter has
to be very heavily doped and the base has to be lightly doped.
This causes the BE
junction capacitance and the base series resistance to be very
high. This problem is
alleviated in HBTs which can have lightly doped emitters and
heavily doped bases.
The output current characteristic of the HBT, according to the
Ebers-Moll model
is given by,
Ic = Ico (exp(qVBEInikT) 1) + Ici (exP(4'Vscln2kT) 1) (2.7)
The output current measurements are usually performed in the
common emitter
configuration in which the collector voltage is swept in the
reverse bias direction for
different base current levels. So, for values of VcE close to
zero both the BE and
BC diodes are forward biased (saturation mode) and the collector
current is low.
With increasing VCE the BC forward bias decreases so that the
collector current
level rises. This happens until the BC diode becomes reverse
biased and the output
current levels out (forward active mode).
Sometimes due to the Early effect a non-zero slope is seen in
the Ic curves in
the forward active region. Since the BE diode is a
heterojunction the cut-in voltages
-
14
for the diodes are different. Thus the diodes start at different
voltage levels and an
offset voltage results in the IC-VcE characteristics of an
HBT.
Gummel plots are used to characterize the collector and base
currents with
respect to the BE diode. In this measurement the BC diode is
shorted and the
voltage across the BE diode is swept. This provides information
about the different
components of the base current in the forward active mode.
Inverse Gummel plots are
obtained by sweeping the BC diode voltage while the BE diode is
shorted [23, 24, 25].
2.5 Single quantum wells and resonant tunnel diodes
If a smaller b.g. material (called well) is sandwiched between
two wider b.g. mate
rials (called barriers) and if the width of the well is
comparable to the de Broglie
wavelength of electrons, the energy levels of the electrons in
the well for motion in
the direction perpendicular to the heterointerface become
discrete. The energy levels
of these quantized states depend upon (1) the well width, (2)
the band offset, (or
the barrier height) and (3) the effective mass of the electrons
in the barrier and well
regions. Thus the energy of the electrons inside the well is a
function of the well
width and the well and barrier materials.
A resonant tunnel diode (RTD) consists of a quantum well
sandwiched between
two thin barrier materials. Heavily doped contact regions are
formed on the other side
of the barrier regions. Fig. 2.4(a) shows the energy levels that
result at equilibrium in
this structure. Conduction through the RTD occurs via tunneling
through the barrier
layers. If the energy of the electrons tunneling though the
barrier layers does not
coincide with the energy level in the quantum well (called the
resonant energy level),
the tunneling probability becomes very low. However, application
of appropriate
bias to the contact regions results in the alignment of the
emitter Fermi level with
the resonant energy level in the well (fig. 2.4 (b)). Then
electrons from the emitter
side of the RTD can propagate through the structure without any
attenuation. This
effect is called resonance. In this situation the tunneling
probability is considerably
enhanced since energy states are available in the well region
for which energy and
-
15
11MM, `WINNE/
Well
Contact Emitter Wellregion Contact
(emitter) region 1=111 11101MM =OM (Collector)
Eo
WM MIMI INN , Ef Ef Eo
1.111111111 Ec Ec
Collector
(b)(a)
Figure 2.4. A double barrier structure in a RTD: (a) at
equilibrium and (b) under biased conditions.
momentum conservation conditions can be satisfied during the
tunneling process.
As the bias is increased further, the resonant energy level gets
misaligned with the
Fermi levels of the contact layers and the current drops. The
device then exhibits
a negative differential resistance characteristic. Multiple
peaks can result due to
resonance with the higher quantized energy levels in the
well.
Barrier symmetry is not essential but a symmetric barrier leads
to interference
of the electron wave function transmitted through and reflected
from two identical
barriers. This interference leads to an increase in the
magnitude of the wavefunction
resulting in more tunneling and greater peak current levels.
However, several factors
like scattering mechanisms within the well and phonon and
impurity assisted tun
neling affect the propagation of electrons through the double
barrier structure. This
renders accurate modelling of the I-V characteristics
difficult.
-
16
However, determining the voltage levels at which the current
peak occurs is
simpler since these are related to the resonant energy levels
within the quantum well
which can be treated as a finite potential well. Under bias
conditions the energy
levels are given by the roots of E in the equation below
[25],
kdo = nir sin-1(I (-yE + 1)E]) sin-1(\1(-yEl[Vo+ VQ + (7
1)E])
(2.8)
Here, E = E, - Ec, k = Ortn(E, Ec)/h, do is the width of the
quantum well, V, is the depth of the potential well under zero-bias
conditions, V, is the voltage
drop across the potential well, -y is the ratio of the effective
mass of an electron in
the barrier to that in the well material and n = 1,2,3 ... are
the integer values for
which solutions exist for the above equation. If the fixed
interface charge density at
the well-barrier interface and the voltage drop across the
quantum well are ignored
and equal voltage drops across the two barriers assumed, the
first resonance peak is
found to occur when the applied voltage equals 2E,/q [25,
26].
-
17
Chapter 3 Physics of Radiation Effects
The study of the effects of nuclear radiation in semiconductor
devices requires a
knowledge about the different radiation environments electronic
systems may be ex
posed to and the damages that result from such exposures. The
damages suffered
by semiconductor devices in a radiation environment depend upon
the types and
energies of the electromagnetic and particle radiations present
in that environment.
The operational lifetimes of space probes and satellites are
determined by the life
expectancies of the on-board electronic systems that have to
operate in the radiation
filled environments present in the Solar System. Nuclear
reactors and situations aris
ing out of nuclear weapon detonation are other special
conditions in which semicon
ductor devices are expected to operate reliably. This chapter
gives a brief overview
of the different types of radiation and radiation environments
modern electronic sys
tems may get exposed to. It then develops in detail the damages
created by such
radiation sources in semiconductor materials. These damages
result in the forma
tion of defects and charges inside these materials. A discussion
about the different
measures that characterize radiation effects concludes this
chapter.
3.1 Radioactivity and types of radiation
Elements with atomic number greater than 82 are unstable and
they spontaneously
emit sub-atomic particles and electromagnetic radiation until
they attain a stable
nuclear configuration. This phenomenon is called radioactivity.
The SI unit for ra
dioactivity is Bequerel, defined as one disintegration per
second. Another parameter
of interest is half life, defined as the time taken by the
radioactive material to decay
to half its original concentration.
-
18
The primary types of radiation resulting from radioactivity are
alpha and beta
particles and gamma radiation. Nuclear reactions like fission
result in the generation
of neutrons. This study covered radiation affects due to all the
above except alpha
particles which have extremely short ranges and can be stopped
very easily. These
are described below:
Neutron: A neutron's mass is identical to that of a proton but
is bereft of
any charge. Due to its large momentum it can cause extensive
displacement
damage in the absorbing material.
Beta: A 0-particle can be positively or negatively charged and
has mass identi
cal to that of an electron. Due to their lighter mass they can
penetrate deeper
into a material but are deflected more easily.
Gamma: It is a very short wavelength electromagnetic radiation
that is highly
penetrating.
The unit for describing particle radiation is flux
(particles/cm's). Time integral
of flux gives fluence. Also the energy spectrum of radiation is
another important
characterizing measure. These constitute what is known as the
intrinsic description
of radiation [1, 2}.
3.2 Radiation environments
The following are the different radiation environments in which
a semiconductor
device may be expected to operate:
Space: This environment consists of high energy particles
trapped by the earth's
magnetic field or those that are present in the Solar System.
They can be cosmic
rays (energetic heavy ions), solar flares (protons, alpha
particles, heavy ions
and electrons) or trapped radiation (mainly electrons in the so
called radiation
belt). The reliability of electronic systems on space crafts and
satellites are the
primary cause of concern in this environment.
-
19
Radiation Processing: Radiation sources are used for the
irradiation of food
products, sterilization of dressing and hypodermic needles and
materials mod
ification. In this environment gamma is the principal source of
radiation.
Weapons: Radiations resulting from nuclear detonation include
neutrons, x-
rays, gamma rays, alpha and /3 particles. This environment
presents a very
large burst of energetic particles and electromagnetic radiation
for which elec
tronic systems need to be hardened for.
Nuclear reactors: In nuclear fission power plants gamma and
neutron are the
primary sources of radiation. Operational reliability due to
exposure to these
radiation sources over a prolonged period of time is the cause
of concern in this
environment.
3.3 Radiation damages in semiconductor devices
The damage generated by a radiation source in a semiconductor
device or material
manifests itself through (1) the displacement of host atoms from
their lattice sites and
(2) the ionization of atoms or internal changes in energy. These
damages result from
the interactions and scattering mechanisms through which the
energetic particles or
photons lose their energy in the absorbing material. A measure
of the amount of
radiation absorbed by a material is rad which is defined as the
exposure received by
a material when 100 ergs of energy has been deposited per gram
of it. The SI unit
of deposited energy is gray which is equal to 1 J /kg of
material.
Atomic displacements occur when an atom gets shifted from its
equilibrium posi
tion in a crystal lattice. An atomic displacement can result
only from the interaction
of an atom with a particle radiation (charged or neutral).
Secondary electrons gen
erated by gamma radiation can also cause this form of damage.
These displacements
can result in defects like vacancies, interstitials,
di-vacancies, vacancy clusters and
impurity-vacancy complexes.
-
20
These defects result in energy levels within the band gap of the
semiconductor
that can trap the free carriers in the material. Mid-gap traps
interact with both con
duction and valence bands and serve as generation and
recombination centers. This
reduces the minority carrier lifetime and increases the leakage
current in a junction
device. Long term issues like stability and reliability arise
from the interaction of
the shallow trap levels with the energy bands.
A consequence of atomic displacements is carrier removal. Also
the defect clus
ters generated by particle damage act as scattering and trapping
centers for the free
carriers. The mobility of the carriers reduces due to these
scattering mechanisms.
The carrier removal rate and mobility degradation are related to
the particle fluence
by,
N No(1 130157,) (3.1) 1
= 1
(1 + ,07,) (3.2) tt
The interaction between the electronic structure of the
semiconductor atoms and
electromagnetic radiation can result in free electrons by any of
the three mechanisms:
(1) photoelectric effect, (2) Compton effect, and (3) pair
production. Charged parti
cle radiation like electrons can also give rise to ionization
effects. Secondary electrons
generated by the above processes if endowed with sufficient
energy can ionize atoms.
Primary effects caused by ionization are the production of
excess charged car
riers and the generation of surface states, charges in
insulators and local electric
and magnetic fields. These electric fields and trapped charges
modify the potential
distribution in the devices affecting current flow through them
[1, 2].
3.4 Measures of radiation effects
Different radiation environments present different levels of
radiation exposure. Nu
clear detonation subjects an electronic system to a burst of
high energy particles
and electromagnetic radiation for a very short duration of time
while the effect of
environments like space and nuclear reactors takes place over a
prolonged period.
Effects due to radiation exposure is thus grouped into four
categories:
-
21
Total dose effect: The absorption of ionizing radiation in a
material of interest
is characterized by the total dose effect. It serves as a
measure of the amount of
radiation a device can sustain before it fails to operate
reliably. Different radia
tion sources produce different total dose effects since the
damage caused varies
from one radiation source to another. This study covered the
total dose effects
due to gamma, neutron and electron radiation in compound
semiconductor
devices.
Dose rate effect: Weapon or nuclear reactor environments can
subject electronic
systems to bursts of radiation of very short pulses. These
transient radiation
effects are used to characterize damages caused by the time
variation of the
photon flux. Possible effects are photocurrent generation and
latchup.
Single event effect: This effect is caused by charge collection
at sensitive nodes
caused by particle radiation. This phenomenon covers both single
event upset
(SEU) and latchup. The former can result in the change of state
of a bistable
logic element.
Neutron effects: This effect is characterized by the total
fluence of neutrons
that a device can sustain before it fails to operate reliably.
Since the energy
of the neutron flux dictates the extent of damage, the fluence
is specified at a
particular energy level.
-
22
Chapter 4 Radiation Study of III-V Interfaces
The operation of heterostructure devices like HEMTs, HBTs and
RTDs involves the
interplay of several physical mechanisms. Isolating the effect
of radiation on these
devices requires a knowledge of the manner in which the
components constituting
these devices get affected by radiation. Single heterojunctions
and single quantum
wells form the integral components of the heterostructure
devices studied in this
research. Hence, preliminary experiments were directed towards
understanding the
degradation phenomenon in each case. At the heterointerface,
non-idealities in the
nature of the non-abrupt transition from one material to another
results in interface
traps and charges. An increase in these charge densities due to
ionizing electromag
netic radiation like gamma can degrade the 2-DEG channel
mobility in a HEMT.
Also radiation induced diffusion of dopants across the
heterointerface can result in
an effective change in the quantum well width. Changes in the
conduction band
offset can affect the quantum well barrier height. These changes
in turn affect the
current levels in a RTD.
This experiment involved studying the integrity of III-V
interfaces under gamma
radiation conditions. Capacitance-voltage measurements were
performed to obtain
the carrier profiles in the structures while photoluminescence
studies were done to
estimate the change in the occupancy levels of the energy states
in the quantum
wells. Data from these experiments were used to calculate
theoretically the change
in some of the material properties.
This chapter begins with the details about the sample
structures, the experi
mental procedures and the measurement set-ups that were used. An
overview of the
theoretical modelling technique follows. The results and
discussions covering this
radiation study on single heterojunctions and quantum wells are
then presented.
-
23
Schottky contacts
E E o 3GaAs: Si0 GaAs: Si kr)CI en n=5e16/cm
CN1
Ci n=8e16/cm 3 ci,
I
heterojunction 1
E Al GaAs:Si In GaAs 100 A Quantuminterface tr)
0
n=2e17/cm3 E GaAs: Si 3 Welloc:3 In n=5e16/cm c::;
n+ GaAs n+ GaAs substrate substrate
Ohmic contacts
(a) (b)
Figure 4.1. (a) Single heterojunction and (b) single quantum
well structures.
4.1 Experimental work
4.1.1 Sample preparation
Fig. 4.1 gives the cross-sections of the heterostructures
studied. Both structures
were grown on n+ substrates using the in-house Perkin Elmer
Model 425B Molecular
Beam Epitaxy system. For C-V measurements the indium on the
bottom surface
of the samples was used as an ohmic contact. Schottky contacts
were obtained by
evaporating gold dots (0.5 mm diameter) to a thickness of 1500
A. The sample size
was about 1 cm x 1 cm for the C-V measurements while for the PL
studies the
sample size was approximately 3 mm x 3 mm.
-
24
4.1.2 Measurement set-up
Photoluminescence measurements:
The measurement set-up consisted of an Argon ion laser (American
Laser Corp.)
with a peak emission of 40 mW at 488 nm. The samples were
mounted on a cold
finger using optical grease and placed in a cyroshroud which had
glass ports for laser
access. The cryoshroud was evacuated to prevent condensation
formation and was
cooled down to 24 K using a closed cycle helium refrigeration
system (Air Products).
The laser was focused on the sample using an optical rail
consisting of mirrors
and lenses. The emission from the sample was focused into a 0.5
m monochromator
(Jarrel-Ash) through an optical filter. The monochromator was
connected to a pho
tomultiplier for light detection. The photomultiplier was cooled
using liquid nitrogen
to reduce thermal dark noise. Standard lock-in detection
technique was used to col
lect the data which was read out through a computer interface
which also controlled
the range of sweep of the grating monochromator. Operating
conditions included
a laser power density of about 1 W/cm2 and a 1100 V bias on the
photomultiplier
tube.
PL measurements use an optical source whose emission energy is
more than the
band-gap of the material under study. This optical source
illuminates the material
and generates electron-hole pairs with energy more than the band
gap of the material.
These optically generated electron-hole pairs then recombine
emitting photons whose
energy is equal to the energy difference between the initial and
the final states of
the electrons. In a quantum well, the energy levels of the
carriers become discretized
and radiative transitions between these discrete energy levels
can be studied.
By plotting the radiative luminescence versus energy,
information about the band
structure and the energy levels present in the material or the
well can be determined.
Since the presence of deep defects affects the intensity of the
radiative luminescence,
PL curves serve as a powerful tool to understand the defect
densities in a quantum
well. Also the roughness of the interface affects the well width
which shifts the
luminescence peaks.
-
25
Capacitance-Voltage measurements:
C-V measurements were performed using a HP-4280A CV meter
operated through
a computer interface. The procedure was automated to yield
capacitance and con
ductance data. These values were used to calculate the carrier
concentration profile
and depletion width using the following formulae,
2n(x) =
EGaAs eoq(Area)2 (dC-2 I dV) (4.1)
eGaAs foArea (4.2)
For simplicity the dielectric constant was assumed to be the
same as that of GaAs
for all the layers.
In the C-V measurements increasing reverse bias was applied to
the Schottky
contact so as to sweep the depletion region through the entire
width of the device.
As the depletion width increased the value of the carrier
concentration could be
obtained from eqn. 4.1. The breakdown voltage of the Schottky
junction had to be
larger than the maximum voltage required to sweep the
layers.
4.1.3 Theoretical modelling
In an ideal heterojunction the interface is smooth and abrupt
with no defects. But in
real heterostructures, small deviations from the abrupt nature
of the heterojunction
over a single atomic plane result in defects at the
heterointerface. Some of these
defects may be charged resulting in the generation of a charge
density (o-ti) at the
interface. Change in this interface charge density serves as a
measure of the amount
of charge generated by gamma radiation at the interface. The
measured carrier
concentration profiles are used to estimate the amount of
interface charge at the
heterojunction using a technique developed by Kroemer et al
[27], in which ai is
fdetermined by,
00
(Nd(x) n(x))dx (4.3)
Here, n(x) is the measured carrier profile obtained from the C-V
measurements and
Nd(x) is the donor charge distribution known from the structure
growth parameters.
-
26
Conserving moments of the carrier, the built-in potential across
the heterojunc
tion is determined by,
Vbi = q f [Nd(x) n(x)](x xi)dx (4.4) 0
The heterojunction conduction band discontinuity is then
obtained from,
AEG, -= qVbi + 6.2 (4.5)
Here, 61 and 82 are the Fermi energies below the conduction
bands in GaAs and
AlGaAs respectively and are determined from the known doping
concentrations in
the two layers.
Carrier profiles in a quantum well structure obtained through
C-V measure
ments can also be used to estimate the band offset at the
well-barrier interface using
the above procedure. However use of the above equations requires
small values of
cri. For lattice matched systems like GaAs/AlGaAs quantum wells
this is true but
for pseudomorphic systems like GaAs/InGaAs quantum wells the
deviations at the
interface may be significantly high to cause large values of the
fixed charge den
sity. This situation precludes the use of these equations and
theoretically generated
carrier profiles with az and AEc as fitting parameters are
matched with the exper
imentally obtained carrier concentration profiles. This yields
values of ai and A.Ec
for the quantum wells. The simulated profiles are obtained from
the self-consistent
solutions of Poisson's and Schriklinger's equations at the
heterointerface [28].
4.2 Results and discussion
4.2.1 Single heterojunction
Fig. 4.2 gives the carrier concentration profile versus the
depletion width obtained
from the C-V measurements of the n-N heterojunction shown in
fig. 4.1. The charac
teristics are for different doses of radiation. The shape of the
curves remains almost
the same up to 40 MRad of radiation with a small shift in the
peak concentration
values. The modelling procedure explained earlier was used to
extract the values of
-
27
x1017
2.4
2.2
Unradiated2
1 MRad 10 MRad
C3
4OMRadci 1.8 a.
c.iz 1.6 O CC
1.4 aC
1.2
1
0.8
0.6 01 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26
DISTANCE (urn)
Figure 4.2. Measured carrier concentration profiles for the n-N
heterojunction.
the fixed interface charge density and the conduction band
discontinuity. Table 4.1
lists the calculated values.
The fixed charge density was found to have increased from -1 x
1011 cm' to
-2 x 1011 cm-2. However the built-in potential and hence, the
conduction band
discontinuity remained approximately the same. From the results
obtained it can
be summarized that since the carrier profiles and band
discontinuity values undergo
negligible change even after a gamma radiation dosage of 40
MRad, this radiation
source does not cause significant damage to the
heterojunction.
The mechanisms by which gamma radiation could have caused damage
are (1)
through the generation of secondary Compton electrons which
would have caused the
heterointerface to become rough and (2) by the process of
radiation induced diffusion
of defects generated at the surface towards the interface. The
fact that the interface
is sufficiently removed from the surface (2200 A) can explain
the small damage
-
28
Dose (MRad) Interface charge (ai, x 1011 cm-2) Vbi (V) AEG
(eV)
0 -1.0 0.165 0.15
1 -1.5 0.163 0.148
10 -1.7 0.162 0.147
40 -2.0 0.167 0.151
Table 4.1. Results of Kroemer's analysis after three doses of
radiation.
suffered due to the second mechanism. In a later experiment
where the effect of the
second mechanism was reduced considerably by using a layer of
PECVD grown oxide,
it was found that identical doses of gamma radiation did not
generate significant
degradation in the mobility values in the 2-DEG in a HFET. Since
mobility in the
2-DEG channel is a sensitive function of the interface quality
and charge density,
it would seem that the first mechanism does not have much effect
either. So, it
can be concluded that for an interface sufficiently removed from
the surface gamma
radiation does not affect its integrity in terms of the
interface smoothness and charge
density.
4.2.2 Single quantum well
Fig. 4.3 shows the carrier concentration profiles obtained from
the C-V measurements
of a single quantum well after various doses of gamma radiation.
Curve fitting
was performed with o-i and DEC as fitting parameters to match
the theoretically
generated curves with the measured ones. For the unradiated
profile, the best fit
was obtained for a fixed charge density of 2.5 x 1011 cm-2 and a
graded composition
of the In GaAs quantum well with the indium mole fraction
varying from 0.1 at
the interface closer to the surface to 0.07 at the interface
closer to the substrate.
The asymmetric distribution of the indium is a result of its
segregation towards the
surface during MBE growth [29].
-
29
O MRad 10 MRad 40 MRadg5
00 Fe 4 cc
00 cc
3 Cl)
2
1
0.15 0.2 0.25 0.3 0.35 0.4 DISTANCE (urn)
Figure 4.3. Measured carrier concentration profile for the
single quantum well.
The carrier profile peaks obtained after radiation show marginal
change. This
suggests that the change in the fixed charge density is small.
The profiles appear
to be more symmetric after radiation suggesting indium
redistribution during the
radiation process. Drevinsky et al [30] have shown that in
silicon enhanced diffusion
of defects occurs during radiation due to the phenomenon of
recombination induced
defect random walk. They suggested that during radiation excess
electron-hole pairs
are generated which recombine to release sufficient energy
necessary to initiate the
movement of defects. In a similar manner the indium that had
segregated near the
surface during MBE growth might have redistributed during
radiation. The concen
tration gradient that results from the MBE growth aids this
process of redistribution.
Fig. 4.4 shows the PL curves from the quantum well obtained
before and after
different doses of radiation. The curves appear to have shifted
lower in energy and
have broadened slightly. This can be due to the fact that the
indium redistribu
-
0.8
C
30
Dose 0 Dose 1 Dose 2 Dose 3
0.6
0.4
0.2
0 I I t 1 3 1.31 1.32 1.33 1.35 1.36 1.371.34
Energy (eV)
Figure 4.4. PL curves for the GaAs/InGaAs SQW after three doses
of radiation. Dose 0 = Pre-rad; Dose 1 = 1 MRad; Dose 2= 10 MRad;
Dose 3 = 40 MRad.
tion process smears the heterointerface causing the effective
well width to increase.
Also the defects close to the interface can assist in the
movement of the indium
atoms across the interface resulting in a small increase in the
well width. A shift
in an outward direction in the carrier concentration profile
(fig. 4.3) supports this
argument.
It would thus seem that single heterojunctions and single
quantum well struc
tures do not undergo significant damage due to gamma radiation
even up to a dose
of 40 MRad in terms of interface smoothness and charge
densities. Some evidence of
the quantum well widening is seen.
-
31
Chapter 5 Radiation Study of Transport in HEMTs
Extremely high values of mobility in the conducting channel
makes HEMTs very
attractive for use in high speed circuits. However 2-DEG
transport in the channel is
a very sensitive function of factors like scattering mechanisms
and concentration of
carriers in the channel. Radiation induced effects like
displacement and ionization
damages increase the former and decrease the latter. This
reduces the overall channel
conductance thereby degrading the performance of the HEMT.
Neutron and gamma radiation studies were performed to identify
and estimate
the radiation damages that affect 2-DEG transport in a HEMT. In
the gamma radia
tion experiment, the effectiveness of an oxide passivation layer
in reducing radiation
damage was studied. It was seen that the passivated samples
suffered less dam
age compared to the unpassivated ones. Neutron radiation
resulted in more severe
degradation in mobility and sheet carrier concentration values.
For this part theoret
ical modelling was performed to estimate the roles played by the
different scattering
mechanisms.
The first part of this chapter deals with the sample preparation
procedure and
the measurement technique used to obtain mobility and sheet
carrier concentration
values. An explanation of the theoretical modelling that was
used to identify the
scattering mechanisms involved is then given. Gamma radiation
results are presented
along with a discussion about the differences in the degradation
seen in the passi
vated and unpassivated samples. Neutron radiation results
follow. Results from the
photoluminescence studies and the theoretical modelling that
were performed are
then presented. A summary of the observations made in this
experiment concludes
this chapter.
-
32
5.1 Experimental work
5.1.1 Sample preparation
The structures were grown using the in-house MBE system at a
substrate tempera
ture of 585 C and an Al mole fraction of 0.2. After completion
of the growth the
wafers were removed and cleaved into 4 mm x 4 mm pieces.
Samples for gamma radiation:
Samples that were used for the gamma radiation study were lapped
to remove the
indium from the backside of the wafer. Ohmic contacts were made
at the four cor
ners of the samples using an ultra fine tipped Antex soldering
iron and an alloy of
10 % Zn/90 % In for p-type contacts and an alloy of 10 % Sn/90 %
In for n-type
contacts. The contacts were then annealed in a forming gas (90 %
N2/10 % H2)
environment for 3 minutes at 450 C. One of the objectives of
this experiment was
to study the effect of surface passivation in reducing radiation
damage. For this pur
pose two samples were prepared from each structure. One had a
clear unpassivated
surface with the ohmic contacts. On the other sample 2500 A of
silicon dioxide was
deposited using Plasma Enhanced Chemical Vapor Deposition
(PECVD) after the
ohmic contacts had been formed. A substrate temperature of 225 C
and a depo
sition rate of 200 A/min were used for the PECVD process. Hall
measurements
performed before and after the growth indicated that the oxide
layer did not affect
the ohmic contacts.
Samples for neutron radiation:
Neutron radiation involved four different levels of fluence.
Since the same piece could
not be radiated over and over again, different pieces cut from
the same sample were
used for each dose. Thus, four pieces were prepared from each
sample (structure).
They were distinguished by making gentle indentations on the
surface. This avoided
-
33
Magnet plates Computer Interface Keith ley Model 181
Nanovoltmeter
Switch Box
LN2 Dewar Keith ley Model 220 Current Source
Figure 5.1. Measurement set-up for Hall measurements.
the need for lapping off of the indium present on the bottom of
the wafer. The
presence of indium increased the mechanical strength of the
samples and prevented
them from breaking due to handling during the radiation and
measurement processes.
Contacts were formed using the alloys mentioned earlier and
subsequently annealed
for ohmic contact formation. Measurements were performed to
obtain pre-radiation
values of mobility and carrier concentration. Samples for
different fluence levels
were packed in separate plastic vials for the neutron radiation
exposure. In the
first run of this experiment most of the samples broke due to
handling. In the next
run, the samples were packed in fiber glass wool inside the
vials. Samples for the
photoluminescence (PL) measurements were obtained by cleaving 3
mm x 3 mm
pieces from the wafer. For each structure four pieces were taken
and indentations
were made to distinguish them.
5.1.2 Measurement set-up
The Hall measurement set-up (fig. 5.1) that was used to perform
the transport stud
ies consisted of an electromagnet, a specially designed sample
holder that could be
placed vertically between the plates of the magnet and a small
Dewar that was
-
34
4 mm
I
4 mm
Figure 5.2. Contact configuration in a typical Hall sample.
used to immerse the sample holder in liquid nitrogen for low
temperature measure
ments. A Keithley Model 220 Programmable Current source and a
Keithley Model
181 Nanovoltmeter were used to perform the measurements. A
computer interface
controlled the measurement and data collection process over a
National Instruments
GP-IB interface. A specially designed switch box was used to
connect the current
source and the voltmeter to the different contacts on the
sample.
5.1.3 Measurements and data collection
Fig. 5.2 illustrates a typical contact configuration on a sample
used for the Hall
measurement using the Van der Pauw method [31]. This method was
used since
it allowed the determination of mobility and sheet carrier
concentration of an arbi
trarily shaped sample. Resistivity of the sample was determined
by passing current
through contacts (A,B) and then measuring the voltage drop
between contacts (C,D).
The measurement was repeated by sending current of the opposite
polarity through
contacts (A,B). The voltage reading from each measurement was
then averaged to
eliminate any offset voltage that might have existed. This
process was repeated three
more times by passing current through the contacts (A,C), (C,D)
and (B,D). The
results were then averaged over all four measurements to obtain
the resistivity of the
sample.
-
35
The next step in the measurement process involved passing of a
current of known
magnitude through (A,D) and measuring the voltage drop across
(B,C) and then
repeating the process by passing the same current through (B,C)
and measuring
the voltage drop across (A,D). The average of these voltage
drops was then taken.
When this measurement was performed in a magnetic field applied
perpendicular
to the sample surface the increase in the voltage drop
represented the Hall voltage,
which is the voltage that is developed across the sample to
counter the Lorentz force
acting on the carriers due to the applied magnetic field.
This measurement was performed in the absence of the magnetic
field and in
the presence of positive and negative magnetic fields (3200
Gauss). The electric field
that generated the Hall voltage was proportional to the applied
current and magnetic
field densities and the proportionality constant, called the
Hall coefficient, gave the
sheet carrier concentration of the 2-DEG. The Hall coefficient
and the resistivity of
the sample were used to determine the mobility of the
carriers.
The mobility and sheet carrier concentration measurements were
performed on
the un-radiated and radiated samples on separate sample holders
to prevent radioac
tive contamination. For the PL studies on the HEMT structures,
the set-up described
in the earlier chapter was used.
5.1.4 Theoretical modelling
This section will briefly cover the physical aspects of the
scattering mechanisms that
affect transport in the 2-DEG channel and the principle behind
the simulation pro
gram that was used in this experiment to estimate the role
played by these scattering
processes in modulating the channel mobility values.
In an ideal crystal with a perfect periodic crystal potential,
an electron moves
with a constant velocity. However defects in the crystal or
thermal vibrations of
the atoms perturb the crystal potential thereby scattering the
electrons from their
path. Defects, impurities and donors within the crystal scatter
the electrons through
Coulombic and neutral scattering mechanisms depending on their
charge states.
-
36
The propagating vibrational modes of the atoms in the crystal
lattice (phonons)
can be classified into four different groups: longitudinal
acoustic, longitudinal op
tic, transverse acoustic and transverse optic. In the acoustic
vibrational mode the
scattering mechanisms that come into play are deformation
scattering potential and
piezoelectric scattering (for polar crystals like GaAs) while in
the optic vibrational
mode, deformation scattering potential and polar-optic
scattering are the principal
scattering mechanisms. Since the amplitude of vibration of the
atoms increases with
temperature, the scattering introduced by atomic vibrations
decreases at low temper
atures while Coulombic and neutral scattering by the impurities
and donors increase
at low temperatures.
Under the relaxation time approximation, it is assumed that if
the action of the
applied forces that tends to change the electron distribution
function is removed the
scattering processes tend to restore equilibrium with a
characteristic time constant
called the momentum relaxation time (Tm). A scattering process
can be defined
quantum mechanically by its matrix element which is used in a
collision integral to
relate the scattering process to rni. The relation between the
matrix element for a
particular scattering process (Hu, ) and its momentum relaxation
time is,
1 Ns Vrtz*2'0= H 12 sinO(1 cosO)d8 (5.1)kkTm, 27r h4 0
The matrix elements for the basic scattering processes like
ionized and neutral impu
rity scattering, acoustic phonon scattering, polar-optic
scattering and deformation
potential scattering that are of interest for transport in a
2-DEG are well charac
terized. Hence, the momentum relaxation time for these
scattering processes can be
obtained using the above equation.
If these scattering mechanisms can be assumed to operate
independently the
total scattering due to all these processes can be obtained by
adding the individual
scattering rates. From eq. 5.1 it is evident that the net
momentum relaxation time
at a particular energy will then be given by the sum of the
reciprocal of individual
momentum relaxation times at that energy.
1 1E (5.2) Tnet TZ
-
37
The mobility of the electrons in the 2-DEG is then given by,
q < Tnet >= (5.3)
171*
where, r(E)E(OfolaE) dE
< Tnet >= (5.4)E(afolaE) dE The simulation program used
the above procedure to calculate values of mobil
ity in the presence of different scattering mechanisms. These
calculated values were
matched with the experimentally measured values to identify the
scattering mecha
nisms. The variables that the program needed are the impurity
concentrations in the
doped AlGaAs (supply) layer, undoped AlGaAs (spacer) layer and
undoped GaAs
(buffer) layer, thickness of the spacer and supply layers,
temperature and number
of electrons in the channel. The sheet carrier concentration in
the channel was ob
tained from a self-consistent numerical solution of one
dimensional Poisson's and
Schrodinger's equations [22, 32, 33].
5.2 Results and discussion
5.2.1 Gamma radiation
As mentioned earlier this experiment involved gamma radiation
study of passivated
and unpassivated HEMT structures. Fig. 5.3 (a) gives the
mobility values obtained
after different doses of gamma radiation on an unpassivated
modulation doped struc
ture at 300 K and 77 K. Fig. 5.3 (b) gives the carrier
concentration of the same
sample. Fig. 5.4 (a) gives the mobility values at 300 K and 77 K
of a sample from
the same wafer whose surface had been passivated by SiO2. and
fig. 5.4 (b) gives its
carrier concentration values after the different doses of
radiation. Fig. 5.5 (b) shows
the MODFET structure that was studied.
From fig. 5.3 it is seen that while the mobility decreases by
about 20 % after a
dose of 40 MRad the carrier concentration values remain almost
the same. However
for the passivated sample neither the mobility nor the carrier
concentration undergo
-
38
106 1013
300K * * 300K 0 77K 0 077K
U
10"
E
lo'
leo 10" 10 20 30 40 0 10 20 30 40
Dose (MRad) Dose (MRad)
Figure 5.3. Measured (a) mobility, (b) carrier concentration
values obtained from an unpassivated modulation doped structure
after four doses of radiation.
any significant change after the radiation doses (fig. 5.4).
This effect was seen across
a number of samples studied in this experiment. Since Hall
measurements performed
before and after the PECVD growth had indicated that the
passivation layer growth
did not affect the transport properties of the carriers in the
channel, the deposition
process would not have resulted in the difference in the
degradation seen in both
samples. The slight difference in the pre-radiated values of
mobility seen in figs 5.3(a)
and 5.4 (a) is due to the fact that the samples were taken from
different parts of the
same wafer and not due to the oxide growth. So, it would seem
that the presence of
the 2500 A thick passivation layer helps to protect the 2-DEG in
those samples.
Mobility of the carriers in the channel is a very sensitive
function of the interface
smoothness and the presence of defects close to the interface on
either side. The
primary mechanism by which gamma radiation induces damage is
through the gen
eration of Compton electrons caused by the scattering of the
gamma ray photons in
-
39
106 io13
300K * * 300K o o 77K 0 077K
0NC
0 _ca
2
lo'
1030 loll 10 20 30 40 0 10 20 30 40
Dose (MRad) Dose (MRad)
Figure 5.4. Measured (a) mobility, (b) carrier concentration
values obtained from a passivated modulation doped structure after
four doses of radiation.
the top layers of the structure. These secondary electrons can
cause displacement or
ionization damage. However 2500 A thick PECVD oxide is not thick
enough to stop
the highly penetrating gamma rays. So, the Compton electron
generation would be
identical for both the passivated and unpassivated samples and
the damages created
by these electrons should be identical for both. However the
mobility and carrier
concentration values of the passivated samples remained almost
unaffected. This
suggests that some other mechanism is responsible for the
degradation seen in the
unpassivated samples.
The reason behind the degradation seen in the unpassivated
samples is believed
to be due to the presence of a free surface close to the
channel. A semiconductor
surface is a veritable source of defects due to the bonding
discontinuities present
at the surface. During the radiation process the defects that
are generated at the
surface can become mobile. As mentioned in the last chapter,
electron-hole pairs
-
40
Sample # Structure Mobility (cm2/Vs) Sheet carrier conc.
(cm-2)
300 K 77 K 300 K 77 K
1-18-3-96 Uniformly doped 7000 130000 6 x 1011 6 x 1011
1-3-2-97 6-doped 5000 115000 7 x 1011 6.7 x 1011
Table 5.1. Typical pre-radiation values for the neutron
samples.
generated during radiation while recombining release sufficient
energy necessary for
the movement of defects through the structure [30]. So, during
radiation the defects
generated at the surface move towards the interface before
freezing in. The presence
of an increased number of defects close to the channel results
in the degradation
seen in the mobility values in the unpassivated samples. The
short distance of the
interface from the surface exacerbates this situation.
However, once the defects penetrate deeper into the structure,
i.e. into the un
doped buffer layer before freezing in, their effect diminishes.
That explains why the
degradation in the mobility values seem to saturate for higher
dose levels.
5.2.2 Neutron radiation
Uniformly doped and delta-doped structures were studied in this
experiment (fig. 5.5).
Fig. 5.6 gives the plots of the mobility and sheet carrier
concentration in the 2-DEG
channel for an uniformly doped sample (# 1-18-3-96). The values
after radiation
have been normalized with respect to their pre-radiation values.
Fig. 5.7 gives the
similar results for the delta doped sample (# 1-3-2-97). Table
5.1 lists the typical
pre-radiation values of